BLS AI exposure categories: which jobs, and what it means at 16

With its 2025-35 projections, the US Bureau of Labor Statistics started rating how exposed each occupation is to AI. Here's what the four categories say, and what they don't.

The BLS AI exposure categories sort 831 US occupations into four groups, Low, Moderate, High and Very high, by how much of the work AI could do or already helps with.^[2][3] They aren't a list of jobs about to disappear. BLS says it plainly: "exposure does not imply job loss."^[2]

What are the BLS AI exposure categories?

The US Bureau of Labor Statistics introduced them on August 27, 2026, alongside its employment projections for 2025 to 2035.^[1] Each occupation gets two scores. Theoretical exposure asks whether AI could assist with or complete some of the work, based on three research studies. Observed exposure comes from two measures of how people actually use AI tools, matched to the tasks of each job.^[2] BLS combined the two and grouped every occupation into one of the four categories.^[2]

At the Very high end, a larger share of a job's tasks can be completed or assisted by AI. At the Low end, the job's requirements don't match current AI capabilities well.^[2] The groups are almost the same size: 213 occupations are Low and 206 sit in each of the other three.^[3] That's because the scale is relative. It ranks jobs against each other, not against a fixed bar.^[2]

Which jobs are in which category?

Here are eleven occupations from the BLS table, with their projected change over the decade and their RIASEC interest code from O*NET, the US Department of Labor's occupation database.^[5]

Look at the Very high rows. Data scientists are among the fastest-growing jobs in the country, while customer service representatives are projected to shrink.^[3] Same category, opposite directions. BLS expects AI to support demand for some occupations while productivity gains dampen it for others.^[1]

Does high exposure mean the job will disappear?

No. BLS writes that a High or Very high rating doesn't necessarily mean employment will fall, and a Low rating doesn't mean a job will be untouched. The category isn't a replacement estimate, and it doesn't separate AI doing the work from AI helping someone do it.^[2] By our own calculation from the BLS data (total 2025 and 2035 employment in each group; BLS doesn't publish this figure), the Very high group as a whole is still projected to grow, by about 2 percent, against about 6 percent for the Moderate group.^[3]

Two more limits are worth knowing. The theoretical scores reflect what AI could do by mid-2023 at the latest.^[2] And BLS calls the uncertainty about AI's impact "very high."^[4]

Does your RIASEC type matter?

Look across these eleven rows and a pattern shows up. The Very high jobs lead with Investigative, Artistic or Conventional letters: research, creating, and organizing information.^{[5][6][7][8][9]} The Low and Moderate jobs lead with Realistic or Social, hands-on work and work with people, plus one Enterprising kitchen job.^{[12][13][14][15][16]} BLS also names arts, design and media, and office and administrative support, as areas where AI may hold back demand.^[1]

That's eleven jobs, not a law. Data scientists are Investigative and Very high, and they're still growing fast.^[3][6]

What this means for you

If you're 16, don't drop an interest because its jobs score Very high. Use the rating to ask better questions.

• Look at both columns. Exposure and projected growth are separate numbers, and a Very high job can still be growing fast, as data scientists show.^[3]
• Ask which tasks AI touches. Open the O*NET page of a job you like and read the task list. Which ones would you still want to do yourself?
• If your code leads with Artistic or Conventional, as the Very high jobs in this sample do, check the rating of each job you're considering, not just the field. If you're Realistic or Social, notice that the hands-on and people jobs in this sample sit lower, then check your own jobs one by one.
• Treat it as a first edition. The categories are new, and BLS calls the uncertainty about AI's impact very high.^[4]

Your next step

Don't know your letters yet? Take the CAT test: 23 questions, about 10 minutes, free. Then read what RIASEC is and check where your jobs sit in the table above. CAT supports career reflection. It's guidance, not a definitive answer.

← All news